Thank Enrico, awesome work! I wanted to suggest to drop maven build on branch 3.4. We never really wanted to release or even migrate maven to 3.4, but for easier backports, we did. Even though maven is there, we said we will just keep releasing 3.4 with ant for backward compatibility reasons. Plus, maven for some reason has some flaw in the unit tests (ZOOKEEPER-3284).
I don't see any added value running maven build on PRs on the 3.4 branch. Any thoughts, should we vote on this? Regards, Norbert On Wed, Jun 12, 2019 at 10:42 AM Enrico Olivelli <eolive...@gmail.com> wrote: > Patrick > I took a look to Yetus, > https://yetus.apache.org/documentation/0.10.0/precommit-basic/ > and it is very cool. > > I will make experiments but it will take time. > > I am going to double check my list of stuff to migrate and if I am done I > will disable the ant based precommit only for master branch. > > We will have: > - travis + jenkins maven on master > - travis + jenkins ant+maven on branch 3.4 and 3.5 > > Enrico > > > Il lun 3 giu 2019, 10:53 Enrico Olivelli <eolive...@gmail.com> ha scritto: > > > > > > > Il giorno lun 3 giu 2019 alle ore 10:48 Norbert Kalmar > > <nkal...@cloudera.com.invalid> ha scritto: > > > >> Hi Enrico, > >> > >> Thanks for looking into this. > >> I also agree on the points to be dropped. Patrick has a valid point on > the > >> contrib tests though. (Maybe a vote on dropping some contrib projects > >> first? :) ) > >> > > > > > >> > >> I'd also like to add I get a lot of javadoc warnings on maven, I'm not > >> sure > >> why Ant doesn't complain, guess different configuration? > >> > > > > I have sent this PR for -Xdoclint > > https://github.com/apache/zookeeper/pull/965 > > > > The build is failing on ASF CI and on Travis with an impossible error, > > that refers to an invalid line. > > I have tested the build on several machines / JDKs and I cannot reproduce > > the error locally > > > > Any help is welcome ! > > Enrico > > > > > >> > >> > >> Regards, > >> Norbert > >> > >> > >> On Fri, May 31, 2019 at 6:04 PM Patrick Hunt <ph...@apache.org> wrote: > >> > >> > Sounds reasonable. Have you looked at what Yetus provides? iiuc a > >> number of > >> > the hadoop ecosystem projects use it to provide base/consistent > >> experience > >> > wrt the build. Now that we're on Maven perhaps we should re-evaluate? > >> > > >> > re #8 - I don't think we should compromise given contrib is still not > >> > decided. > >> > > >> > Patrick > >> > > >> > On Fri, May 31, 2019 at 8:39 AM Enrico Olivelli <eolive...@gmail.com> > >> > wrote: > >> > > >> > > Hi, > >> > > you can find here (1) the current ANT-based precommit script that > runs > >> > > againsts every Github Pull Request . > >> > > > >> > > These are the steps of the script and how we are doing the same in > >> Maven: > >> > > 1) check for @author tags -> handled with maven-checkstyle-plugin > >> > > 2) Check for new tests in the patch -> not handled with maven > >> > > 3) check javadoc warnings -> not handled with maven (we should add > >> > > -Xdoclint in Maven) > >> > > 4) check java compilation for warnings -> handled with -Werror > >> > > -Xlint:deprecation -Xlint:unchecked (this is slightly different, in > >> ANT > >> > we > >> > > have a generic -Xlint and we were counting for the number of > warnings) > >> > > 5) check for "releaseaudit" -> handled with "Apache Rat Plugin" in > >> maven > >> > > 6) check "findbugs" -> handled with "spotbugs" with maven > >> > > 7) run "tests" -> handle with maven default build "verify" > >> > > 8) run "contrib tests" -> not handled with maven > >> > > 9) add JIRA comment -> not handled with maven (not needed IMHO, we > >> have > >> > > github PR status lines) > >> > > > >> > > My questions to the community: > >> > > - I would like to drop 2), "check for new tests": I think it gives > no > >> so > >> > > much value, and it is tricky to reimplement, but I can do > >> > > - I would like to drop 8) "run contrib tests": there is an open > >> > discussion > >> > > about dropping "contrib" at all > >> > > - I would like to drop 9) -> add jira comment: We already have a > >> better > >> > > workflow with github, no need to spam JIRA. I can see a little value > >> in > >> > > having some track in JIRA that "the patch was good", but I don't > >> think it > >> > > is worth to continue to put JIRA credentials on jenkins (from a > >> security > >> > > perspective) > >> > > > >> > > Please take a look and give your opinion, if no one objects and some > >> > > committer supports my changes we will be able to drop ANT precommit > >> soon > >> > > > >> > > Regards > >> > > > >> > > Enrico > >> > > > >> > > [1] > >> > > > >> > > > >> > > >> > https://github.com/apache/zookeeper/blob/master/zookeeper-server/src/test/resources/test-github-pr.sh > >> > > > >> > > >> > > >
